1. Field of the Invention
This invention relates to a video signal coding method and a decoding method, and more particularly, applicable is to the case where feature points of an image are detected and then high-efficiency coding of the video signal is performed.
2. Description of the Related Art
Heretofore, as a method for high-efficiency coding of a video signal, there has been a method wherein an input video signal is serial-parallel converted by means of a DCT (discrete cosine transform), and then adaptive quantization according to a human visual property is performed for each frequency band, or a method wherein an image is subband-divided using wavelet base and then weighted so as to code for each band. According to these with methods, distortion is not particularly visible and a high compression rate can be obtained.
However, in these coding methods, such a disadvantage exists that a visual undesirable effect such as a block distortion becomes noticeable as a compression rate is raised. Then, as a coding method which can suppress a visual undesirable distortion under a high compression rate, there is a structure extracting and coding method according to change point detection of an image, in which the distinctive points of the structure of the image are extracted and effectively coded.
For example, as shown in FIG. 1, in a structure extracting and coding apparatus 1 which codes a video signal using a structure extracting and coding method, an input video signal S1 is quantized by a quantizer 2, thereby a quantization coefficient S2 is generated, which is sent to a multiplexer 3.
On the other hand, the structure extracting and coding apparatus 1 inputs the input video signal S1 to a two-dimensional change point detecting circuit 4, which detects the feature points, and sends a flag "1" to a chain coder 5 as a feature point signal S3 if the present signal is detected as the feature point, otherwise sends a flag 0. A chain coder 5 chain codes the pixels of which feature point signals S3 are "1" in one frame, and sends it to the multiplexer 3 as a chain code output S4.
The multiplexer 3 multiplexes the chain code output S4 and the quantization coefficient S2 at the feature point, and then sends this to a buffer memory 6 as a multiplexed feature point signal S5. The buffer memory 6 smooths an information quantity of the multiplexed feature point signal S5, and outputs it as an output signal S6 of the structure extracting and coding apparatus 1 according to feature point detection of the image.
Here, the two-dimensional change point detecting circuit 4 is constructed, for instance, as shown in FIG. 2. More specifically, the two-dimensional change point detecting circuit 4 once stores the input video signal S1 in a frame buffer 10, and then smooths it by means of a filter 11. A smoothed signal S10 output from the filter 11 is sent to a lateral direction processing block 12 and a longitudinal direction processing block 13, respectively.
The lateral direction processing block 12 supplies the smoothed signal S10 to an adder 16 via a delay element 14 and a multiplier 15 of which multiplication coefficient is -1 respectively, thereby a difference between the present signal and the signal preceding one sample is found by the adder 16, so that a lateral direction amplitude inclination signal S11 is generated. The lateral direction amplitude inclination signal S11 is squared by a multiplier 17, thereby a lateral direction amplitude inclination signal power S12 is obtained. The lateral direction amplitude inclination signal power S12 is supplied to a multiplier 21 via a delay element 18 and an adder 19 to which an inverted threshold value T is given, and also supplied to the multiplier 21 via an adder 20 to which the inverted threshold value T is given. Thereby, in the multiplier 21, a difference signal S13 between the lateral direction amplitude inclination signal electric power S12 and the threshold value T is multiplied by a difference signal S14 between the one sample preceding lateral direction amplitude inclination signal power S12 and the threshold value T, and the multiplication result signal S15 is sent to a comparator 22. The comparator 22 sends a flag "1" in the case where the multiplication result signal S15 is less than 0, or a flag "0" in the case where it is "0" or more to an OR circuit 23 as a lateral direction change point detection signal S16.
The longitudinal direction processing block 13 supplies the smoothed signal S10 to an adder 26 via a line delay 24 and a multiplier 25 respectively, thereby a difference between the present signal and the signal preceding one line is obtained by the adder 26, so that a longitudinal direction amplitude inclination signal S17 is generated. The longitudinal direction amplitude inclination signal S17 is squared by a multiplier 27, thereby a longitudinal direction amplitude inclination signal power S18 is obtained. The longitudinal direction amplitude inclination signal power S18 is supplied to a multiplier 31 via a line delay 28 and an adder 29 to which an inverted threshold value T is given, and also supplied to the multiplier 31 via an adder 30 to which the inverted threshold value T is given. Thereby, in the multiplier 31, a difference signal S19 between the longitudinal direction amplitude inclination signal power S18 and the threshold value T is multiplied by a difference signal S20 between the longitudinal direction amplitude inclination signal power S18 preceding one line and the threshold value T, and the multiplication result signal S21 is sent to a comparator 32. The comparator 32 sends a flag "1" in the case where the multiplication result signal S21 is less than "0" or a flag "0" in the case where it is "0" or more to the OR circuit 23 as a longitudinal direction change point detection signal S22.
By ORing the lateral direction change point detection signal S16 with the longitudinal direction change point detection signal S22, the OR circuit 23 outputs a two-dimensional change point detecting circuit output signal S23.
In coding of feature point coordinates based on the conventional two-dimensional change point detecting method and chain coding described above, with respect to the change points of the linear differential value power of the input video signal against a certain threshold value, for instance the points at which a sharp change arises locally, the two-dimensional change point detecting circuit 4 at first detects two points of the starting point and the ending point of that change as feature points, and the following chain coder 5 separately chain codes.
Therefore, in the conventional feature point coordinates coding method, on the edge portion at which a sharp change arises, the position coordinates of the two curves are always coded, hence such a disadvantage has existed that the code quantity increases correspondingly.